[llvm-commits] [llvm] r60314 - /llvm/trunk/lib/Transforms/Scalar/GVN.cpp

[Nick Lewycky]

Chris, this patch breaks Obsequi on x86 Linux. Unfortunately, I don't 
have a testcase for you because bugpoint-{opt,llvm-ld,llc,cbe} all just 
reduce down to a nullary testcase.

Unfortunately, I can't build a current tree with just this patch backed 
out. Please investigate, and let me know if you need any testing done on 
a linux box.

Nick

Chris Lattner wrote:
> Author: lattner
> Date: Sun Nov 30 19:31:36 2008
> New Revision: 60314
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=60314&view=rev
> Log:
> Make GVN be more intelligent about redundant load 
> elimination: when finding dependent load/stores, realize that
> they are the same if aliasing claims must alias instead of relying
> on the pointers to be exactly equal.  This makes load elimination
> more aggressive.  For example, on 403.gcc, we had:
> 
> <     68 gvn    - Number of instructions PRE'd
> < 152718 gvn    - Number of instructions deleted
> <  49699 gvn    - Number of loads deleted
> <   6153 memdep - Number of dirty cached non-local responses
> < 169336 memdep - Number of fully cached non-local responses
> < 162428 memdep - Number of uncached non-local responses
> 
> now we have:
> 
>>     64 gvn    - Number of instructions PRE'd
>> 153623 gvn    - Number of instructions deleted
>>  49856 gvn    - Number of loads deleted
>>   5022 memdep - Number of dirty cached non-local responses
>> 159030 memdep - Number of fully cached non-local responses
>> 162443 memdep - Number of uncached non-local responses
> 
> That's an extra 157 loads deleted and extra 905 other instructions nuked.
> 
> This slows down GVN very slightly, from 3.91 to 3.96s.
> 
> 
> Modified:
>     llvm/trunk/lib/Transforms/Scalar/GVN.cpp
> 
> Modified: llvm/trunk/lib/Transforms/Scalar/GVN.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/GVN.cpp?rev=60314&r1=60313&r2=60314&view=diff
> 
> ==============================================================================
> --- llvm/trunk/lib/Transforms/Scalar/GVN.cpp (original)
> +++ llvm/trunk/lib/Transforms/Scalar/GVN.cpp Sun Nov 30 19:31:36 2008
> @@ -915,11 +915,28 @@
>      }
>    
>      if (StoreInst* S = dyn_cast<StoreInst>(DepInfo.getInst())) {
> -      if (S->getPointerOperand() != L->getPointerOperand())
> +      // Reject loads and stores that are to the same address but are of 
> +      // different types.
> +      // NOTE: 403.gcc does have this case (e.g. in readonly_fields_p) because
> +      // of bitfield access, it would be interesting to optimize for it at some
> +      // point.
> +      if (S->getOperand(0)->getType() != L->getType())
> +        return false;
> +      
> +      if (S->getPointerOperand() != L->getPointerOperand() &&
> +          VN.getAliasAnalysis()->alias(S->getPointerOperand(), 1,
> +                                       L->getPointerOperand(), 1)
> +            != AliasAnalysis::MustAlias)
>          return false;
>        repl[DepBB] = S->getOperand(0);
>      } else if (LoadInst* LD = dyn_cast<LoadInst>(DepInfo.getInst())) {
> -      if (LD->getPointerOperand() != L->getPointerOperand())
> +      if (LD->getType() != L->getType())
> +        return false;
> +      
> +      if (LD->getPointerOperand() != L->getPointerOperand() &&
> +          VN.getAliasAnalysis()->alias(LD->getPointerOperand(), 1,
> +                                       L->getPointerOperand(), 1)
> +            != AliasAnalysis::MustAlias)
>          return false;
>        repl[DepBB] = LD;
>      } else {
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>